I think the question is not about performance, but about usability.
Generally, it's worthwhile to implement a good search engine than doing what you're proposing.
Committed Usability
Imagine a scroll bar 1000 pixels high ( almost the height of a 1600 x 1200 monitor ). This means that for every pointable pixel, there will be 20.000.000 / 1.000
records = > % with% records per pixel.
It turns out that by scrolling through the logs, you will go to the first of that 20,000 sequence, you will always show how many will fit on the screen, and the rest of the 20000 is hidden ... here comes the question: How are you going to do to access the rest of the 20,000 records? With the arrow on the keyboard? With the mouse scroll wheel? Page down and page up keys. This is unfeasible. What's more, hunting for pixels on the scrollbar is already somewhat counterproductive to start the conversation.
Alternative:
User is usually like this, does not know what he wants ... he thinks: I need access to any record, so I put all the records on the same screen. He does not think, I need access to any record, so I need an effective way to reach them, so I need a search engine that meets my needs.